We show that the shift in the transition temperature of the dilute homogeneous Bose gas is nonanalytic in the scattering amplitude a. The first correction beyond the positive linear shift in a is negative and of order a(2)lna. This nonuniversal nonanalytic structure indicates how the discrepancies between numerical calculations at finite a can be reconciled with calculations of the limit a-->0, since the linearity is apparent only for anomalously small a.
We show that the shift in the transition temperature of the dilute homogeneous Bose gas is nonanalytic in the scattering amplitude a. The first correction beyond the positive linear shift in a is negative and of order a(2)lna. This nonuniversal nonanalytic structure indicates how the discrepancies between numerical calculations at finite a can be reconciled with calculations of the limit a-->0, since the linearity is apparent only for anomalously small a.
Beyond standard approaches in existing literature, we explore the relative shift in the transition temperature of a homogeneous dilute Bose gas using the Cornwall-Jackiw-Tomboulis effective action formalism within the improved Hartree-Fock approximation. The first correction to the enhancement of the transition temperature, as compared to that of the ideal Bose gas, is expressed in a universal form: it is linear with respect to the scattering length. The slope of this linear relationship shows excellent agreement with the exact numerical calculations presented in previous studies. Additionally, we identify non-universal terms contributing to the shift.
We apply the nonperturbative optimized linear $\ensuremath{\delta}$ expansion method to the O(N) scalar field model in three dimensions to determine the transition temperature of a dilute homogeneous Bose gas. Our results show that the shift of the transition temperature $\ensuremath{\Delta}{T}_{c}{/T}_{c}$ of the interacting model, compared with the ideal-gas transition temperature, really behaves as $\ensuremath{\gamma}{\mathrm{an}}^{1/3}$ where a is the s-wave scattering length and n is the number density. For N=2 our calculations yield the value $\ensuremath{\gamma}=3.059.$
We discuss certain specific features of the calculation of the critical temperature of a dilute repulsive Bose gas. Interactions modify the critical temperature in two different ways. First, for gases in traps, temperature shifts are introduced by a change of the density profile, arising itself from a modification of the equation of state of the gas (reduced compressibility); these shifts can be calculated simply within mean field theory. Second, even in the absence of a trapping potential (homogeneous gas in a box), temperature shifts are introduced by the interactions; they arise from the correlations introduced in the gas, and thus lie inherently beyond mean field theory – in fact, their evaluation requires more elaborate, non-perturbative, calculations. One illustration of this non-perturbative character is provided by the solution of self-consistent equations, which relate together non-linearly the various energy shifts of the single particle levels k . These equations predict that repulsive interactions shift the critical temperature (at constant density) by an amount which is positive, and simply proportional to the scattering length a ; nevertheless, the numerical coefficient is difficult to compute. Physically, the increase of the temperature can be interpreted in terms of the reduced density fluctuations introduced by the repulsive interactions, which facilitate the propagation of large exchange cycles across the sample.
Within a quasiparticle framework, we reconsider the issue of computing the Bose-Einstein condensation temperature (Tc) in a weakly non-ideal Bose gas. The main result of this and previous investigations is that Tc increases with the scattering length a, with the leading dependence being either linear or log-linear in a. The calculation of Tc reduces to that of computing the excitation spectrum near the transition. We report two approaches to regularizing the infrared divergence at the transition point. One leads to a a∣lna∣-like shift in Tc, and the other allows numerical calculations for the shift.
We calculate explicitly the variation δTc of the Bose-Einstein condensation temperature Tc induced by weak repulsive two-body interactions to leading order in the interaction strength. As shown earlier by general arguments, δTc/Tc is linear in the dimensionless product an1/3 to leading order, where n is the density and a the scattering length. This result is non-perturbative, and a direct perturbative calculation of the amplitude is impossible due to infrared divergences familiar from the study of the superfluid helium lambda transition. Therefore we introduce here another standard expansion scheme, generalizing the initial model which depends on one complex field to one depending on N real fields, and calculating the temperature shift at leading order for large N. The result is explicit and finite. The reliability of the result depends on the relevance of the large N expansion to the situation N=2, which can in principle be checked by systematic higher order calculations. The large N result agrees remarkably well with recent numerical simulations.
The leading-order effect of interactions on a homogeneous Bose gas is theoretically predicted to shift the critical temperature by an amount $\ensuremath{\Delta}{T}_{c}\ensuremath{\simeq}{\mathrm{ca}}_{\mathrm{sc}}{n}^{1/3}{T}_{0}$ from the ideal gas result T0, where asc is the scattering length, n is the density, and c is a pure number. There have been several different theoretical estimates for c. We claim to settle the issue by measuring the numerical coefficient in a lattice simulation of O(2) ${\ensuremath{\varphi}}^{4}$ field theory in three dimensions---an effective theory which, as observed previously in the literature, can be systematically matched to the dilute Bose gas problem to reproduce nonuniversal quantities such as the critical temperature. We find $c\phantom{\rule{0ex}{0ex}}=\phantom{\rule{0ex}{0ex}}1.32\ifmmode\pm\else\textpm\fi{}0.02$.